PXA-bench-files-GGUF

PXA benchmark files

Competitor-side model files used in the PXA cross-engine comparisons (bench/fair-battle.md), published so every row can be reproduced from a download. Each is a stock model converted with the repo's own llama-quantize.

Full sums in checksums.sha256. Protocol: bench/fair/protocol.md in the engine repo.

Measured artifacts behind the PXA README rows

The single-card rows on the PXA front page were measured on the two files below and on nothing else. They carry the names they have in bench/fair/weights/MANIFEST.sha256, so a download here is byte-for-byte the file a published number came from.

Both are PXA-codec files: they need the PXA engine, not stock llama.cpp.

The rest of the PXA-Fusion2-35B manifest

bench/fair/weights/MANIFEST.sha256 in the engine repo lists four more Fusion2 artifacts. None of them backs a published single-card speed row — they are here because the manifest names them, and because the recipes in docs/COOKBOOK.md and the SWE-bench Lite reproduction in bench/swe/ call for them. I put them in this repo so the whole manifest resolves from one place instead of half of it pointing at a repo name that now belongs to a different model.

The .gguf tiers above are PXA-codec files: they need the PXA engine, not stock llama.cpp, and they must be fully GPU-resident (-ngl 99).

Check what you downloaded

sha256sum -c checksums.sha256 --ignore-missing

Run it from the directory holding the files — the names in checksums.sha256 are bare. The same sums are in bench/fair/weights/MANIFEST.sha256 in the engine repo, and bench/fair/run.sh refuses to start if a file present on disk does not match.

I keep these here because the repo they were first published in, poisonxa/PXA-Fusion2-35B-GGUF, was renamed to poisonxa/PXA-Fusion4-35B-GGUF — a different model on its main branch, though the original bytes are still reachable there at revision 2d7d9e7e5f445b1ea38a5686b8cf9d6bc9aa29df. An old link now lands on the rename, so the copies in this repo are the ones to use: they carry the manifest's own filenames and the same hashes, and they are the files the published numbers were measured on.

Changes

• 2026-10-24: PXA-Fusion2-35B-PXQ6-MTP.gguf was replaced (new sha256 573f6c74424721944a8e40c2cb37de744ef24df59ec021106f4eeeb0dc0c4b44, same size). The three expert tensors of the MTP layer (blk.40) were stored in an old format that current PXA builds refuse to load; they are now stored as MXFP4 with the same values (every weight dequantizes to exactly the same number as before). Every other tensor and all other metadata are unchanged, except one metadata field (quantize.imatrix.dataset) that held a build-machine path and now holds a generic one. The file as it was before this change stays downloadable at revision d4d8f0e88e97dc75c1cfc7e3b3399aba5aff39d0. Until the next PXA release, bench/fair/weights/MANIFEST.sha256 in the engine repo still lists the old sum for this file; checksums.sha256 here is current.